Research Exam 3

Type 1 and Type II Error

(not on study guide) - Not possible to absolutely say that null hypothesis true or not true, only probably true or probably not true - we make statistical inferences based on incomplete info-so there is always a risk of error - Type I: reject the null when it is true - Type II: accept the null when it is false

Bivariate descriptive statistics

(not on study guide) describes the relationship between two variables. Two types: Crostabulations: A two-dimensional frequency distribution in which the frequencies of two variables are crostabulated. Ex: Do men smoke more heavily than women or vice versa. Correlation: relationships between two variables can be described by correlation methods. The question is "to what extent are two variables related to each other?" Ex: to what degree are anxiety scores and blood pressure values related.

Concepts Essential to Health Services Research

- Designed to document the quality and effectiveness of healthcare and nursing services - Often focuses on parts of a health care quality model developed by Donabedian: Structure of care (nsg skill mix) Processes (clinical decision making) Outcomes (end results of pt care) - reflect structure (staff/edu), process (satisfied/assess/intervene), and outcomes of nursing care (quantity/quality) (ANA) - access, quality, and cost - major issues: organize/finance, access, practitioner/pt/consumer behavior, quality, clinical eval/outcome research, informatics/clinical decision, health professions workforce - every discipline involved because every action affects another discipline - does HSR shape public policy or does public policy shape HSR? both impact each other

Mix methods

- Research that integrates quant/qual data and strategies in a single study or coordinated cluster of studies - many areas of inquiry can be enriched by triangulating this data, some require mixed methods: pragmatism paradigm (positivist = quant, constructivist = qual) - Advantages: complementarity (avoid limitations for just using one), practicality (complex), enhanced validity (support/greater confidence) - Reasons for use: Start out with Qual then move to quant, need both to answer, helps better analyze secondary data, quant can be difficult to interpret on own, helps to meet objectives - Applications: instrument develop, intervention, hypothesis, theory, explication (see below) - types that involve intervention: clinical trials, evaluation research, nsg intervention research Explication- qualitatively explaining the meaning of quantitative descriptions or relationships QUAN + QUAL = concurrent; both types happening at the same time (convergent parallel) QUAL → quan = sequential; qualitative happens first; collected in phases (exploratory) QUAN → qual = sequential; quantitative happens first; collected in phases (explanatory) QUAL/quan means qual is dominant status, if QUAL/QUAN means codominant status Convergent parallel design: obtain diff, but complementary data about the central phenomenon under study i.e. to triangulate data sources Explanatory design: sequential designs with quantitative data collected in the first phase, followed by qualitative data collection in the second phase Exploratory design: sequential MM design, with qualitative data being collected first

Representative sampling

- a sample whose key characteristics closely approximate those of the population- a sampling goal in quantitative research - more easily achieved with: probability sampling, homogeneous populations, larger samples achieved through power analysis (how many you need in your study to determine statistical significance) - *usually based on ROL, other studies done, support of evidence in lit, involves statistician; not done in qualitative*

Outcomes research

- a subset of health services research - comprises efforts to understand the end results of particular health care practices and to assess the effectiveness of health care services - represents a response to the increasing demand from policy makers and the public to justify care practices in terms of improved patient outcomes and costs - provides evidence about benefits, risks, and results of treatment

Measures of variability

- concerned with spread of data; how spread out are the values in a distribution - answer the question "is the sample homogenous or heterogeneous? - is the sample similar or different? *want them to be similar* - means can be the same but differ in ___ describe how much dispersion in sample 1. Range (difference between highest and lowest score) - range is___ - Simplest & most unstable measure of variability - based on only two most extreme scores - indicates total spread of scores - mean deceiving with these two sets of scores - outliers: far outside range of majority Ex: Test scores ranged from 64-95 → range = 31 2. Percentile: percentage a given score exceeds - data point below which lies a certain % of values in a frequency distribution - Ex: score in the 90th percentile is exceeded by 10% of scores *not same as percentage* 3. Standard deviation → measures average deviation of scores from mean; reported w/ mean, based on concept of normal curve - most frequent measure of variability - larger deviation from mean, the larger the SD - the larger the SD is the more variation in the population response - care for all patients using critical pathways to determine "average"

Statistical tests and significance

- goal is probability - decide test based on level of measurement, type of hypothesis, difference between groups or relationship between 2 or more variables, sample size, how many groups, observations or scores: dependent or independent, how many observations or scores - significant value means relationship or difference probably not caused by chance

Research critique key areas

- must be able to evaluate in order to use it, distinguish between best-evidence Thorough examination of all parts of research study: First, read entire article, make initial evaluation Next, evaluate each part in depth Focus on Design - before you see how research actually carries out, can see if other part are congruent Ex: pretest - posttest - should see two groups in problem, hypothesis, sample - often no right or wrong, be objective, give RATIONALE - what, why, how, clarity, significance, str/weakness?

Critiquing sampling plans: considerations

- the type of sampling approach used (convenience, consecutive, random) - the population and eligibility criteria for sample selection - the sample size with a rationale - a description of the sample's main characteristics (age, gender, clinical status, and so on)

Metasynthesis

- theoretical integration and interpretation of qualitative findings - diverse and complex, not a literature review - frequency effect sizes, themes, integrating findings on experience - represents a family of methodological approaches to developing new knowledge based on rigorous analysis of existing qualitative research findings - the bringing together and breaking down of findings, examining them, discovering essential features, and combining phenomena into a transformed whole - integrations that are more than the sum of the parts-novel interpretations of integrated findings - DEBATE: exclude low-quality, integrate studies based on multiple qual traditions, various typology/approaches/terminology - contains no quantitative data, just combined themes from multiple qual studies - steps: formulate problem, decide on design: selection criteria, search strategy, search for data in the literature, evaluation of study quality, extract data for analysis, data analysis and interpretation QUALITATIVE

Factors encouraging questionnaire return-rate Advantages & disadvantages of questionnaires

- time, hand-addressed, personal signature, motivational info, incentive, neat/clear of instrument, ease of completion, time required to complete, guarantee of anonymity, inclusion of a preaddressed, stamped envelope Advantages: quick&generally inexpensive, easy to test for reliability and validity, admin is time efficient, can obtain data from wide-spread geo areas, anonymity can be guaranteed in cover letter Disadvantages: costly to mail, potential low response rate, respondents may provide socially acceptable answers or fail to answer, respondents may not be represented of the pop, no opportunity to clarify items that may be misunderstood, respondents must be literate, respondents must have no prohibitory physical handicap

Clinical trials

- type of intervention research - assess clinical intervention and test innovative therapy/drug in phases - studies that develop clinical interventions and test their efficacy and effectiveness - undertaken to evaluate an innovative therapy or drug or often designed in a series of phases - emphasis on EBP led to call for trials, practical/pragmatic clinical trials used in making real-world applications phase I: establish safety tolerance and dose; focus is develop the best tx; usually a simple design with no control group. Usually one group pre/post test phase II: seeks preliminary evidence of effectiveness, PILOT TEST using QUASI experimental; look for possible side effects and identify possible refinements; small scale or quasi experimental Phase III: full experimental test of efficacy of tx. RCT with random assignment to tx conditions under control, often in multiple sites; objective= develop evidence about efficacy. Clinical trial most frequently refers to this phase; AKA efficacy study Phase IV: study of the effectiveness of intervention in the general population; emphasizes external validity of an intervention in the general population, emphasis on generalizability

Frequency distribution

- used to calculate rates of occurrence, provide info for data trends over time - divide freq. of event in given time period by all possible occurrences of event during same time period Ex. divide # surgical site infections in a month by total # surgical procedures in that month # of times each event occurs is counted or data are grouped & frequency of each group reported One of the simplest ways to present data *GIVES A GOOD OVERVIEW* < 20 scores = list all (16: 4, 17: 2 etc.) > 20 scores = may group into intervals (16-18: 8; 10-15: 12) but may lose some significant info that way Frequency-symbol is 'f' 'N' meaning number of participants is also used

Three types of descriptive statistics

1. Frequency distribution: Impose order on numeric data. An arrangement of values from lowest to highest and a count or percentage of how many times each value occurred. Positive skew: personal income (less people have higher incomes in US) Negative skew: age at death (less people die young) Many human attributes have a normal distribution 2. Central tendency: indicate what is "typical". (i.e. mean, median, mode) 3. Variability: Two distributions with identical means could differ with respect to how spread out the data is. This is variability. (i.e. range and standard deviation)

Interval

= rank ordered and placed into categories; distance between groups/ranks can be measured Third level of measurement Actual numbers on a scale (equal distances between points on the scale) - zero is arbitrary Inferential analysis Ex: body temperature, wt, ht, Likert scales, hgb, age, celsius - can calculate mean, median, mode - cannot multiply, divide or calculate ratios due to no true zero

References

Accuracy, completeness Match text citations

Questionnaires

Advantages: quick & generally inexpensive, easy to test for reliability & validity, administration is time efficient, can obtain data from widespread geographical areas, anonymity can be guaranteed w/ cover letter - less costly and are advantages for geographically dispersed samples; offer possibility of anonymity, which may be crucial in obtaining information about certain opinions or traits - one of many data collection methods - BEST method for human response data, paper-pencil format, answers given in writing - development of a reliable and valid questionnaire is difficult, many literature resources available for use in the construction - knowledge levels, opinions, attitudes, beliefs, ideas, feelings, perceptions Validity: - rests on validity of the data obtained, governed by the respondents' willingness or ability to provide accurate info Guidelines: - neat and attractive, minimal length, grammatically correct, no errors, clear margins, high-quality printing/paper - preferred language of respondents', appropriate knowledge and reading level (6th grade), avoid slang, colloquialisms, medical/nsg jargon - determine reading levels: flesch reading ease, fog scale level, flesch-kincaid grade level - questions short, less than 20 words, split long questions up, affirmative not negative manner - avoid ambiguous: more than one interpretations - double negative: never doesn't rain - double barreled: do you eat healthy and exercise Questions: - demographic (gather characteristics about sample) - open-closed ended - contingency (if no, go to question six) - filler (reduce emphasis on other questions, researcher has no direct interest) - group all about same topic together, demographic at begin/end, simple then complex - <10 min to complete, 2-3 pgs or less Cover letter: - written clearly and simple instructions, factor of motivation, all mailed questions - elements: ID of the researcher and any sponsoring agency or person, purpose, how participant selected, reason the respondent should answer; length of time to complete, written clearly, important factor, all mailed, how data will be used or made public; deadline for return; offer to inform respondents of results; researcher's contact info, personal signature Distribution: - convenient location, mailing or distribution system, through internet

Assumptions

All studies based on assumptions Universal "all humans need to feel loved"; Common Sense "all respondents will answer honestly" Should flow from Theoretical Framework Stated and Implied

Theoretical framework

Appropriate to the study Nursing and/or other discipline Concepts clearly defines/ relationship

Overarching questions

Are results valid? Bias? Sound scientific methods? What are the results? Do numbers add up? Will results help me care for my patients? Applicability Who conducted it Qualifications , credentials Expertise in that area Sponsors - any conflict of interest Could you search university's website or author info

Comparison of calculated value vs. critical value

Calculated value comes from the data Critical value is set statistically & the stats program determines for you If the calculated value is ≥ the critical value at a p value of ≤.05 → reject the null b/c findings did not occur by chance, findings statistically significant, lends support to research hypothesis - EX. obtained p value .03 - reject null If the calculated value is < the critical value at a p value of >.05 → fail to reject the null b/c findings did occur by chance, findings not statistically significant, does not lend support to research hypothesis - EX. obtained p value .3 - fail to reject null

Introduction

Catch interest Briefly explore study area Significance; statistics

Critique key areas

Clarity Significance What was done; Why was done; How was done ALL studies Strengths and weaknesses Substandard if... partial information; unusable, can't replicate study based on info given; doesn't answer questions asked

Title

Clear, concise Brief - not >15 words Focus is apparent; key words Population, and major variables Avoid extraneous words - "A Study of.."

Instruments

Clearly described Scoring procedures; range of poss. Scores Reliability and validity Use in previous studies

Problem statement

Clearly identified Usually one sentence Too broad? Too many variables?

Lit review

Concise, comprehensive Relevance of sources Support and Oppose expected results Classics , then current sources Paraphrase, not long quotes Show how this study will add to body of knowledge

Other considerations

Correct grammar, sentence structure, and punctuation Last question "after reading this research report, would i refer it to a colleague?"

Measures of relationships

Correlation: to what extent are variables related?, measures of relationship (not necessarily cause and effect) - is there relationship btw amt of time spent with pt & # of requests for pain med by that pt - ordinal or higher level data - scatter plots - visual representations of strength & magnitude of relationship between 2 variables - perfect + and - correlation - range from -1.00 to +1.00 - 0 = absence of any relationship - strength of correlation demonstrated by how close data points approximate straight line (intensity & direction) - closer correlation coefficient to zero = lower/weaker correlation - closer to 1 (+/-) = higher stronger correlation - ***positive correlation: higher score on one variable, higher score on the other - Ex. temp and pulse rate positively correlated = as one incr. so will the other one - ***OR lower score on one variable, lower score on other - Ex. younger age = shorter stay in hospital - Positive correlation (both go in same direction): as anxiety level incr. pulse rates incr. - Ht & Wt: greater Ht tends to be associated with increased Wt - perfect positive correlation would be where tallest person is heaviest - Negative correlation: higher score on one variable, lower score on other measure - inverse relationship: blood volume decrease, see initial rise in pulse rate - anxiety levels incr. = test scores decr. - high self-esteem = low depression - ***CORRELATION DOES NOT = CAUSATION*** - results that only describe or explain cannot be used to directly ID the cause of the findings

Sample

Demographics Measures to protect participants Informed consent/ permission, IRB approval Anonymity; confidentiality Should mention any dropouts that occured

Analysis

Descriptive stats on sample characteristics Scores on variables in study Inferential stats if tested hypothesis Hypothesis supported or not Results of specific tests, degrees of freedom, probability values Clear text and tables Completely objective Data- bound

Sampling/standard error

Discrepancies that inevitably occur when small group (sample) is selected to represent characteristics of larger group (population) - discrepancies that inevitably occur when a small group (sample) is selected to represent characteristics of a larger group (population) - is always some likelihood a sample will NOT accurately reflect the population; findings will be inaccurate - by chance or hypothesis effect? - hypothesis testing: based on rules of "negative inference"

External validity

Discuss limits to generalization Ex; non probability sampling, convenience

Internal validity

Discuss threats History (no control but occurring at same time as study in world), maturation (over time), mortality, selection Bias (unequal group)

Methodologic Studies

Entails investigations of the methods for conducting rigorous research Focuses on development/validation/evaluation of research tools and instruments

Focus of outcomes research (driving forces); What is NSI?

Focuses on the end result of patient care (found in an elsevier online textbook) - In order to explain the end results, nurse researchers also must understand the processes used to provide patient care. - NSI: nsg sensitive indicators = pt outcome solely r/t nsg care received - reflect structure (staff/edu), process (satisfied/assess/intervene), and outcomes of nursing care (quantity/quality) (ANA)

Conclusion

Generalization Data- based, but sometimes may go beyond data Author demonstration meaning and worth Answers "so what"

Systematic reviews

Integration of multiple studies on a topic - umbrella for metasynthesis and meta-analysis - top of evidence hierarchy - the systematic and rigorous integration and synthesis of evidence is a cornerstone of EBP - impossible to develop "best practice" guidelines, protocols, and procedures without organizing and evaluating research evidence through systematic review - narrative, qualitative integration (traditional review of quantitative or qualitative results)

Confidence level and level of significance; means what? Minimum acceptable level of significance

Level of significance = probability of making a Type I error (rejecting a null hypothesis that is, in fact, true) .05 means accepting risk that out of 100 samples, a true null hypothesis would be rejected 5 times OR 5% chance the result really occurred by chance - or 95/100 cases, a true NH would be correctly accepted - 5% occurred by chance, or willing to risk being wrong 5% of the time p= Minimum level of significance in nsg studies is .05 = standard minimum level for alpha in scientific research - a = .01 significance level means 1/100 cases reject true NH (don't put zero in front of decimal) - larger sample size promotes likelihood that extreme cases will cancel out each other - do not prove hypothesis: but "null has high probability of being incorrect" - *rejecting null shows support for scientific hypothesis* - probability of incorrectly rejecting true null - "significance" does not reflect importance or meaningfulness, means results not likely due to chance at some specified level of probability - nonsignificant results: probability high that observed difference or relationship could be result of chance (non reproducible) - diff that researchers with help of statistical inference, consider real, are said to be statistically significant - diff that could easily be caused by chance are not considered real and are nonsignificant; does not mean research hypothesis is not correct - "are or are not significant" Confidence level/confidence interval = probability level in which null hypothesis can be rejected w/ confidence & research hypothesis can be accepted w/ confidence If they have a confidence level of 95%, researchers risk being wrong 5 times out of 100 - probability level in which null hypothesis can be rejected with confidence and research hypothesis can be accepted with confidence - how willing is the researcher to be wrong when declaring that one group really is different from the other group on the variable being measured? Probability: likelihood an event will occur.

Purpose/meaning of standard deviation

Measures average deviation of scores from mean Is reported w/ mean Most frequent measure of validity Larger deviation from mean, larger the SD The larger the SD is the more variation in the population response

Construct validity

Most difficult to measure Concerned w/ what is this instrument really measuring & does it adequately measure the construct of interest Ways to measure → known-groups procedure tests relationships based on theoretical predictions, factor analysis

Review use of outcomes research

Mostly focused on the process-pt-outcomes nexus

Null hypothesis versus research hypothesis-which one is tested; language used r/t the null hypothesis to describe findings. Meaning of rejection of the null

Null hypothesis = no relationship between the variables; only used in statistics, not being tested - statistical, states "no difference", "no relationship" - test this one statistically - if small differences or low correlations found -> chance is reason -> null not being rejected - large diff/correlation not by chance, so null rejected - either accepted (likely due to chance), or rejected (indicate existence of relationship btw variables, unlikely due to chance) Research hypothesis = states the anticipated relationship between the variables; actual hypothesis that researcher is testing - scientific, based on theoretical framework - directional -predict results of study - never tested statistically, never say "proved", not say "rejected" because never actually tested Two explanations: 1. reflects researcher's scientific/research hypothesis 2. diff due to chance factors: reflects null hypothesis, no actual relationship, result of chance or sampling fluctuation Reject the null hypothesis = support the research hypothesis Fail to reject the null hypothesis = does not support the research hypothesis - hypothesis testing: process of disproving or rejection - it is possible to show that the null hypothesis has high probability of being incorrect - such evidence supports scientific hypothesis (research hypothesis) - goal of statistical test is to determine probability

Abstract

Often the only section read Should contain essentials: Hypothesis or Research Questions, methods, description of participants, major findings 100 - 200 words

PICO

P: population I: intervention C: comparison O: outcome

Difference in parameter v. statistic

Parameter: characteristics of a population usually not known - inferential stats: info from sample used to make inferences concerning parameter - generalize findings from sample to population, mean sample stats, population parameter Statistic: characteristics of samples you usually are able to compute (descriptive stats for sample characteristics) You compute statistics to estimate parameters Must describe data in meaningful terms

Results

Present data Data, numbers, descriptive, and inferential statistics Analysis of data Problem, hypothesis - address each one Each test, results (summarization) Narrative and/or table, graphics Objectively presented Do not omit data, even if not significant

Hypothesis

Present tense; relationship between variables Consistent terminology Empirically testable Able to gather objective data Only one prediction

Percentages

Proportion of subgroup to total group IMPORTANT TO REMEMBER → ≥20 values needed for statistics to be valid (20 is minimum) Always report underlying frequencies w/ percentages; if not, can be misleading Round to one or two places (will not always add up to 100% because of rounding) Ex. "60% of students interviewed said scrubs should be required in the classroom as well as in clinical." How many interviewed?

Purpose of inferential statistics; based on what assumption?

Purpose: use data from sample to draw conclusion(s) about population - test hypotheses: how much results fro chance? how stronly are variables associated? Help ID systematic variations btw or among groups within sample? - Based on the assumption that chance is the only explanation for relationships discovered in research - Based on the assumption that researcher used a random sample - assumption that chance is the only explanation for relationships discovered in research - researcher wants to demonstrate that chance is not the reason for relationships found in research - used to determine the likelihood that the sample chosen actually represents the population - helps researcher determine if difference between 2 groups (exp/control) is real difference or only a chance difference that occurred because an unrepresentative sample was chosen from population - larger the difference found, lower probability difference occurred by chance - make an educated guess about a population parameter based on a statistic computed from a sample randomly drawn from that population - take raw data -> understandable - used to test our prediction or hypothesis useful in nursing: - using nonrandom samples decrease ability to generalize results - replication of studies: similar findings; allows conclusions that are similar to conclusions as if one random sample were used

Experimental, Quasi-experimental

Quasi-experimental: trials without randomization in medical literature. May also lack a control group. The signature is an intervention in the absence of randomization Experimental: Contain an experimental and control group, along with randomization. Sometimes called posttest design

Purpose/Aim/Goal

Reasons for undertaking study May give specific goals/ objectives

Relationship of reliability and validity, interrater reliability

Reliability: The extent to which scores for people who have not changed are the same for repeated measurements. Concerned with the absence of variation in measuring a stable attribute for an individual. Validity: a measurement context is the degree to which an instrument is measuring the construct it purports to measure. Close association Reliability is a condition for validity An instrument cannot be valid unless it is reliable Reliability tells nothing about degree of validity An instrument can be very reliable but w/ low validity Interrater reliability: The most typical approach to evaluate how reliable the measurements reflect attributes of the person being rated rather than the attributes of the raters. Involves having two or more observers independently applying the measure with the same people to see if the scores are consistent across raters.

Absolute zero

Remember with ratio, absolute zero, must indicate complete absence of whatever it is measuring. Ask yourself if it is realistic to have absolute zero of the variable being measured; most physical biological measures can have 0 and are measured at ratio level, even if it seems they would be dead

Critiquing research

See syllabus examples: 28-32 (quant), 26-27 (qual) p. 66 in textbook for quant, p. 67 in textbook for qual

Statistical significance v. statistical non-significance

Significance Means results not likely due to chance at some specified level of probability Does not reflect importance or meaningfulness Non-significance Probability high that observed difference or relationship could be result of chance (not reproducible) Does not mean the research hypothesis is not correct Could mean that sample size is too small

Importance of replication of studies

Similar findings Allows researcher to draw conclusions similar to conclusions if one random sample were used

Implications

Specific for: Nursing practice, nursing education, nursing research Possible to replicate Further development of instrument Large sample Author should consider limitations and findings of previous studies

Tables, graphs, figures

Supplement the text Economize/ condense the text Not repeat the text Precise titles, headings Understandable Located with related text Not leave random/ sporadic blank areas within data unless explained

Types of errors

Systematic Error: consistently measures wrong (i.e. measures everyone's wt 2 pounds heavy) Random Error: measuring some samples differently without pattern - randomization is more effective as # of objects/people incr, but is always possibility that incomplete randomization results in the DV rather than the cause (IV) -> DV - statistical inferences tells res. if random error can be ___ as possible explanation for results

Population: targeted and accessible

Target: entire population of interest; entire group of interest based on eligibility criteria Accessible: portion of the target population that is accessible to the researcher, from which a sample is drawn

Research design

Type If experimental - 3 criteria met? (control, randomization, intervention) Tx described Method of Assigning participants Setting described in general terms

Limitations

Uncontrolled variables Research should clearly identify aspects of study over which no control has been exercised Inappropriate tool, weather, small sample

Examples of Measures of Outcomes Used in Outcomes Research

Unruh and Zhang used 9 years of data from 124 hospitals in Florida to examine the relationship between changes in RN staffing and patient safety events Safety, effectiveness, equity, efficiency, timeliness, system responsiveness, pt-centeredness - minority veterans, examine drug Rx, health insurance competition, value of measuring quality and ID barriers of care - John Wennberg = study on geographic variations in medical tech - AHRQ (for national and state statistics) Designs: RCT, cross-sectional, cohort, meta-analysis, systematic reviews Mortality, morbidity, disparity

Biophysiologic methods:

Used for creating independent variables and for measuring dependent variables. Ex. Explores the ways in which nursing actions and interventions affect physiologic outcomes, product assessments. In vivo measurements: performed directly within or on living organisms BP, body temp, HR, RR In vitro measurements: Data is gathered from participants by extracting biophysiologic material from them and subjecting it to analysis by laboratory technicians chemical measures, microbiologic measures, and cytologic or histologic measures. UA, BMP, ABG

Secondary analysis

Uses previously gathered data to address new questions Can be Qn or Ql Cost-effective May not be aware of data quality problems and typically face "if only" issues of inadequate data

Methodology

What, How, Who, Where, When [was the data collected]

Homogenous

When only people who are similar with respect to confounding variables are included in the study. Ex. If age was considered a confounder, participation could be limited to a specified age range.

Experimenter effect

a threat to internal validity in which the experimenter, consciously or unconsciously, affects the results of the study researcher's behaviors/characteristics influence respondents' behaviors (way I ask questions)

Sources of error

all research contains some error - human beings - environmental or setting types of sources of error: - instrument inadequacies, instrument administration biases, environment variations during collection of data, temporary subject characteristics during the collection of data

Mortality/attrition threat

attrition in groups being compared. If different kinds of people remain in the study in one group versus another, then these differences, rather than the independent variable, could account form group differences in outcomes.

Discussion

author interprets, more subjective (more opinionated) Compare findings with those in lit review No new lit review source introduces In light of theoretical framework - findings supportive or not Statistical significance/ clinical significance How limitations may have affected

Data collection in quantitative research

basic decision in use of : 1. new data, collected specifically for research purposes OR 2. existing data (won't be using this type with QUASI experimental) - records (patient charts, hospital records such as ns shift reports, school records student absenteeism, corporate records health insurance choices, letters, diaries, minutes or meetings, photographs), historical data, existing data set (secondary analysis) Major types: 1. self-report- *unless making observations it is self-report data*; structured, data are collected with a formal instrument - interview schedule: questions are prespecified but asked orally, either facetoface or telephone - questionnaire: questions prespecified in written form, to be self-administered by respondents - closed ended (fixed alternative) questions: for example, "were you ever? (yes/no)" - open-ended questions: "why?" 2. pt-reported outcome 3. observation- observe behavior/characteristics 3. biophysiologic measures- clinical info/BP Overview: structure, quantifiability, objectivity

Instrumentation threat

changes in the measurement of the variables or observational techniques that may account for changes in the obtained measurement (ex: different types of thermometers must be calibrated before and after data collection)

chi-square (x^2)

compare frequencies (counts, not means) obtained or observed in categories w/ frequencies expected to occur by chance/if null is true (no relationship) - Assess whether relationship exists between two nominal-level variables (differences in proportions) Nonparametric, nominal data (frequencies or percentages) Ex: Study the relationship of gender & source of HIV Pearson correlation → correlation coefficient designating the magnitude of relationship between two interval- or ratio-level variables Correlation coefficient → Summarizes the degree of relationship between variables ranging from -1 to +1 - if obtained freq. quite diff from expected freq. at specified p-value, null is rejected

Eligibility criteria

criteria designating specific attributes of the target population, by which people are selected for inclusion in a study - the characteristics that define the population: 1. inclusion criteria 2. exclusion criteria

Measures of central tendency

describe average member of sample Answer question "what does the average...?" - summarize the sample, describe average, typical, or most common value for a group of data Mean (average of all scores; interval or ratio; M,X) Most stable & most used measurement of central tendency Larger the sample, less the mean is affected by a single score Very sensitive to outliers or extreme values; pulled in direction of outliers (median and mode not affected by outliers) Most tests of significance use means - all things being equal; choose mean; more powerful statistical tests can be applied to it than to median and mode - exceptions: highly skewed distributions, nominal and ordinal data Mode/modality (most frequent score or value) Only measure appropriate for nominal data; report as % or frequency Crude estimate of average value Sometimes designated 'Mo' Median = the value that falls in the middle - 50% above and below it - best used when data are skewed (less affected by extreme scores) - always somewhere btw mean and mode, arrange in numerical order - designated 'Md, Mdn' Ex: 4 4 5 5 7 8 8 8 9 → median = 7

Define outcomes and health services research and their goal

health services research: broad interdisciplinary field that studies how organizational structures and processes, health technologies, social factors, and personal behaviors affect access to health care, the cost and quality of health care, and ultimately, people's health and well-being - focus on three major aspects of health care: access, quality, and cost - evaluates the effects and outcomes of the health care "system" on people's health - The GOAL: the goal of HSR is to provide info that will eventually lead to improvements in the health of the citizenry (IOM)

T test

parametric examines difference between means of 2 groups of values Useful w/ small (<30) sample Two forms: 1. Independent/Pooled → no association between scores of groups; experimental & control group Ex: compare research course grade (mean score of 92) of SPRING students w/ research course grade (mean score of 88) of FALL students completing research course - *differences between groups, compare group means or averages, if sample means are far enough apart then t-test shows significant diff or statistical significance* 2. Dependent/Correlated/Paired → scores/values are associated Ex: anxiety scores of mothers & daughters compared; each subject in one group matched w/ subject in another group on some variable like age, wt Same group measured pre & post intervention (BP measured pre-exercise & post-exercise) - If calculated t-value (your data) >/= to critical value (table/computer), null is rejected (means that the mean scores are significantly different) *PRIMARY REASON for LACK OF STATISTICAL SIGNIFICANCE IS NOT HAVING LARGE ENOUGH SAMPLE*

In vivo measurements

performed directly within or on living organisms BP, body temp, HR, RR - very accurate and good provision - strong accuracy, objectivity, validity, and precision - may be cost-effective for nurse researchers - BUT caution may be required for their use, and advanced skills may be needed for interpretation

In vitro measurements

performed outside the organism's body (e.g. urinalysis)

sensitivity and specificity

sensitivity: the instrument's ability to correctly ID a "case" i.e. to dx a condition specificity: the instrument's ability to correctly ID noncases, that, is to screen out those without the condition

Meta-analysis

statistical integration of results used to compute common effect size - statistical integration of results used to compute common effect size advantages: objectivity- statistical integration eliminates bias in drawing conclusions when results in different studies are at odds, increased power- reduces the risk of a type II error compared to a single study - despite these advantages, it is not always appropriate; indiscriminate use has led critics to warn against potential abuses steps: 1. problem formulation-delineate research question or hypothesis to be tested 2. design of meta-analysis- identify sampling criteria for studies to be included 3. search for evidence in literature- develop and implement a search strategy (database, unpublished?, avoid publication bias, key words) 4. evaluation of study quality- locate and screen sample of studies meeting criteria (omit low-quality, more high-quality, use of scale/component approach, analyze low and high to see if effects differ = sensitivity analyses) 5. extraction and encoding data for analysis 6. calculation of effects 7. data analysis - Criteria: must decide whether statistical integration is suitable, research question or hypothesis should be essentially identical across studies (avoid apples/oranges), must have a sufficient knowledge base = must be enough studies of acceptable quality, results can be varied by not totally at odds - analytic decisions: effect size index, effect size d, fixed effects model, model, subgroups, quality, publication bias - a key component of meta-analysis is the calculation of an effect size index QUANTITATIVE

Testing threat

the effect of taking a pretest on the participants answers on a posttest

Manipulation

the introduction of an intervention or treatment in an experimental or quasi-experimental study to assess its impact on the dependent (outcome) variable

Setting

the physical location and conditions in which data collection takes place in a study.

Tight controls

used to minimize bias and enhance the interpretability of results

Lab setting

used when the researcher needs to have a great deal of control over the research situation, which often gives greater internal validity to the study

Analysis of Variance (ANOVA)

→ compares differences among >2 means, simultaneously - parametric Examines variance between means, within groups Value you get is an "f" value - 3 or more groups - could use with two groups but results same as with t-test, good to eliminate bias between groups - three or more groups means ANOVA tests significance of variance btw group means, group means of three or more groups - can be used when > one independent variable - Repeated measures (RM-ANOVA)→ 3 or more measures of the same DV for each subject (same indiv. Exposed to intervention A, B, C) OR multiple measures of same DV collected longitudinally at several points in time Ex: effect of humorous distraction on pre-op anxiety...1)20 min recorded comedy routine, 2) 20 mins instrumental music, 3) control/comparison group Ex: Pt BP before, during and after dx exam - prevents from having to do 6 different t-tests between groups so the answer is more accurate

Common projective measures

(not on study guide) - inkblot test, thematic apperception test (usually used with kids)

Delphi technique

(not on study guide) Uses several rounds of questions Seeks a consensus on a particular topic, policy, or subject matter from a group of experts Not necessary to bring experts together in a face-to-face meeting - short-term forecasting

Face validity

"face" Appears to measure desired construct Experts review instrument to validate it Use of the instrument w/ people w/ characteristics similar to study Based on judgement; no objective criteria for assessment

Personality inventories

(can be used to assess personality characteristics) - use self-report measures, assess the differences in personality traits, needs, or values of people Ex. MMPI, EPPS, 16PF also projective techniques: - responses are believed to reflect the subject's internal feelings, subject is asked to: describe the stimuli, tell what the stimuli represents

Q Sort

(not on study guide) Subjects sort statements into categories according to their attitudes toward, or rating of, the statements Subjects are presented w/ a # of words or statements that are written on cards or pieces of paper Subjects sort cards or papers into a # of predetermined categories or piles Forced-choice exercise

Threats to External Validity of study (Based on Methods)

*Interaction of Testing and Treatment* • occurs when a pretest sensitizes participants to the treatment yet to come *Interaction of Selection and Treatment* • a treatment effect only holds true for the specific sample of participants you selected

Ratio

*ends with "o" so there is a zero* = data is categorized & ranked; distance between ranks is equal & there is a "true" or natural zero Highest level of measurement Inferential analysis 0 = total absence of quantity measured - debate usually exists between interval and ratio level Ex: money in bank account, pain of a person, Kelvin, temp, wt, volume, speed ***most physical human measurements are ratio even if it seems impossible to have zero level and be alive***

Nominal

*sounds like "name" you are simply naming categories, dichotomous/categorical* = assignment of objects/events into categories Lowest level of measurement (with no order) Ex: gender, marital status, religious affiliation - demographic questionnaire - MUST BE: mutually exclusive and exhaustive - use descriptive statistics; few statistical tests -> pie charts, mode, range, frequency; no natural order - Ex. gender, marital status, ethnicity, diagnostic, group

Ordinal

*sounds like "order" with natural order" = rank ordered & placed into categories (based on their relative standing on an attribute) Exact differences between rank not possible, zero is arbitrary Can be broken down into nominal data Second level of measurement Ex: mild, moderate, severe (N/V); education levels, military rank - tall, taller, tallest - STILL MUST BE: mutually exclusive & exhaustive - use descriptive statistics; pain scale, pt satisfaction, tumor stage, apgar score; measure attitudes/perceptions - median, percentile, rank

Observational research

- data gathered through visual observation, nurses are well qualified to use this method, carefully developed plan is essential - behaviors observed, who will observe, what procedure, what type of relationship Ex. psycho motor, personal habits, nonverbal, interrater reliability: degree to which two or more raters are observers assign the same rating or score to an observation - range from structured (data collection & expected checklist, observer indicates freq of occurrence) to unstructured (attempts to describe, requires high degree of concentration & attention) - combo: utilizes observation guide, designed with some preconcieved categories, also allows the observer to record additional behaviors, provided both quant/qual data - event (involves observation of entire event Ex. change of shift report) and time sampling (observations of events/behaviors during certain specified times Ex. 2h after dinner)

critiquing data collection process

- find the section, check level of measurement, see if the instruments are described clearly and thoroughly, check to see if a pilot study was performed - questions: did the research report provide info on data collected, was the appropriate level of measurement used, section that described the instrument, describe the instrument thoroughly? used previously? reliability? validity? - read entire research report carefully, try to tell if the best data was obtained from the method, consult research books if not familiar with the methods used

Level of measurement considerations & questions to ask with data collection

- level is appropriate for the type of data desired, degree of precision that is desired for the study - When would you need a high level of precision? when there is a lot of risk/harm involved with a wrong value; use the most precise method = RATIO Appropriate level of measurement: - precision- interval or ratio - ranked or categorized sufficient - ordinal - categories of data only needed- nominal Questions: - what, how, who, where, when? What data will be collected: - type of data needed for the study, examples are data to measure- knowledge, attitudes, behaviors How will the data be collected: - research instrument, major decision for the study, examples are: questionnaire, interview - consider physical stamina of patients Who will collect the data: - researcher only person, team of members, consistency of data collection What is the problem if data is not consistently collected in the same way? invalid, inaccurate, unreliable data Where will the data be collection? - setting for data collection, optimum conditions are considered When will the data be collected? - data collection time: month, days, hours - time period for data collection

Factors affecting data quality in quantitative research

- procedures used to collect the data - circumstances under which data were gathered - adequacy of instruments or scales used to measure constructs - psychometric assessment: evaluates the measure's measurement properties - reliability: extent to which sources are free from measurement error Face validity: whether the instrument looks like it is measuring the target construct Content validity: the extent to which the instrument's content adequately captures the construct Criterion validity: the extent to which the scores on a measure are a good reflection of a "gold standard" - two types: concurrent (comparison) and predictive (ACT) Construct validity: the degree to which evidence about a measure's scores in relation to other variables supports the inference that the construct has been well represented

Data collection factors

- research question(s) or hypothesis(es) - design of the study - amount of knowledge available about the variable(s) Data collection methods: - self-report questionnaires, interviews, physiological measures, attitude scales, psychological tests, observational methods - variety collection: more than one method used; similar results from variety of methods = greater confidence in study findings Selection of data: collection instrument - conduct a lit review, determine measurements on the instrument (reliability = consistency, validity = does it measure what it is supposed to measure) Sources for research instruments: mental measurement yearbooks, instruments for clinical health care research; the measurement of nursing outcomes; health and psychosocial instruments Development of new instruments: - challenging tasks, constructing the instrument, performing a pilot study, considering time, cost, and availability for pilot Data collection criteria: practicality, reliability, validity HINT: projective tests are particularly useful with small children because of their limited vocabularies

Descriptive statistics

- researcher's observations -> data - Data: words or numbers, large amounts of data must be gathered, organized, and summarized - data ALONE do not answer research questions or test hypothesis GOAL: describe sets of numbers & make accurate inference about groups based upon incomplete information - measures to condense data, communicate numerical characteristics of samples (demographics of sample) - Definition: summarize, organize, describe, evaluate, interpret, & communicate various characteristics of sample; frequency of occurrence of concept - Purpose: condense objective observations into usable data Types: 1. central tendency: describe average member of sample 2. variability: describe how much dispersion in sample 3. relationships: magnitude & direction of relationships - may be reported in Text or Tables/Graphs

Observation methods:

- structured observation of prespecified behaviors (involves the use of formal instruments and protocols that dictate what to observe, how long to observe it, and how to record the data) - focus of observation, concealment, method of recording observations Direct observation of people's behavior: Advantages: researchers have flexibility. Can be made through the human senses and then recorded manually, but they can also be recorded digitally. Can be used when participants may not provide reliable answers. Usually designed to capture the behaviors of infants, children, or people whose communication skills are impaired.

Visual Analogue Scale (VAS)

- used to measure subjective experiences (pain, N/V) - measurements are on a straight line measuring 100 mm (have pt put mark and then measure with 100 mm ruler) - end points labeled as extreme limits of sensation Presents subjects w/ a straight line drawn on a piece of paper Line is anchored on each end by words or short phrases Phenomenon extremes are listed at ends of the line Subjects are asked to make a mark on the line at the point that corresponds to their experience of their phenomenon Line usually 100 mm in length; EXACT Quantitative data is obtained from measurements of the responses Useful in measuring: nausea, pain, fatigue, SOB

Practicality of the instrument

-cost -appropriateness -question to consider: length of time for administration? -training to get results -determine before reliability or validity

Representatives of sample is determined how?

A sample whose key characteristics closely approximate those of the population- a sampling goal in quantitative research More easily achieved with: probability sampling, homogeneous populations, larger samples achieved through power analysis Power analysis: statistical formula used to determine sample size needed so results are valid (not by chance); calculated from previous studies/RoL Increase variables = increase sample size

Interviews

Advantages: higher response rate, in-depth responses, wide range of participants, high percentage of usable data, ability to observe verbal & nonverbal behavior - appropriate for more diverse audiences, some people cannot fill out or forget to mail back a questionnaire - opportunities to clarify questions or to determine comprehension - opportunity to collect supplementary data through observation Disadvantages: time consuming, expensive, arrangements may be difficult; participants may be influenced by the interviewer's characteristics, intentionally provide socially acceptable responses, be anxious because answers are being recorded - method of data collection, interviewer obtains responses, face to face encounter or by telephone - purpose: obtain factual data about people, measure opinions, attitudes, and beliefs - schedule: contains a set of questions to be asked by an interviewer - record interview data: entered directly on the interview schedule, recorded on a separate coding sheet, recorded on audiotape, recorded on videotape - interviewer training: responsibility of the study investigator, should be rigorous, should be carried out in groups; provide description of study/purpose/methodology; provide explanation of interview schedule, purpose of each question, meanings of all words, use of probes - unstructured: interviewer directs the course of the interview, conducted like a normal conversation, topics pursued at the discretion of the interviewer; used in exploratory/qualitative research, interviewer may start with broad opening, further questions and probes used, give more freedom in question format, produce more in-depth info, conducted more like a normal conversation, give topics at the discretion of the interviewer - produces more in-depth info, subject's beliefs, attitudes - structured: used a structured set of questions, ask same questions/in same order/in same manner, very objective - semistructured: interviewers ask a certain number of specific questions, additional probes are allowed or encouraged, closed-ended and open-ended question Prior to interview: - intro self, explain purpose, how participant was selected, how obtain info will be used, how long interview will last During: - make comfy, privacy, control noise, use understood language, conversational tone, no right/wrong answers, no pressure, sensitive questions at end After: - ask if they have questions, explain further, thank, compensate, how they can get results Face-to-face influencing factors: ethnic background, age, gender, manner of speaking, manner of dress; telephone = tone and dialect

Maturation threat

Arises from processes occurring as a result of time rather than the independent variable. I.e. wound healing or postop recovery time.

Structured observations

Category systems -> Checklists - formal systems for systematically recording the incidence or frequency of prespecified behaviors or events - systems vary in their exhaustiveness: - Exhaustive system: all behaviors of a specific type recorded, and each behavior are assigned to one mutually exclusive category - Non-exhaustive system: specific behaviors, but not all behaviors, are recorded Rating scales: ratings are on a descriptive continuum, typically bipolar - ratings can occur: at specific intervals, upon the occurrence of certain events, after an observational session (global ratings)

Reliability of measurement (instrument)

Consistency and accuracy that instrument measures the target attribute. Want score 0.80 or greater (1) Stability (of the instrument): Consistency over time, stable & accurate Extent to which scores are similar on two separate administrations of an instrument Evaluated by test-retest reliability Participants complete same instrument on two occasions Appropriated for relatively enduring attributes -things that don't change overtime (i.e. creativity) Ways to measure: Physiological instruments - stable and accurate Questionnaire instruments - test and retesting High correlation coefficient - close to 1.00 (the closer the correlation is to one, the higher the reliability) 0.4 to 0.7 = weak 0.7 to 0.9 = moderate > 0.9 = good Equivalence reliability: Two or more observers use same instrument: Interrater or interobserver reliability - two different forms of instrument: degree to which same results occur, alternate or parallel forms reliability - the degree of similarity between alternative forms of an instrument btw multiple raters/observers using an instrument - most relevant for structural observations - assessed by comparing agreement btw observations or ratings of two or more observers interobserver/interrater reliability Tools for testing reliability: - mechanical device, written questionnaire, human observer - measurement by correlational procedures - reliability assessment involves computing a reliability coefficient range 0-1, less than 0.7 considered unsatisfactory, 0.8 or higher desirable (2) Internal consistency: Individual items measure the same variable One concept or construct (i.e. variable) is measured Sample of items is its main consideration Extent to which all the items on an instrument are measuring the same attribute Alphas greater than or equal to 0.80 highly desirable; accept 0.70 (3) Equivalence: the condition of being equal or equivalent in value, worth, function, etc. The state or fact of being equivalent; equality in value, force, significance, etc

Indirect observation

Data is gathered from participants by extracting biophysiologic material from them and subjecting it to analysis by laboratory technicians chemical measures, microbiologic measures, and cytologic or histologic measures. UA, BMP, ABG

Exhaustive v. Exclusive

Exhaustive system: all behaviors of a specific type recorded, and each behavior is assigned to one mutually exclusive category Ex: recording every action a child takes while playing with another child Nonexhaustive system: specific behaviors, but NOT all behaviors, are recorded. Ex: count every time a child hits another child, but not including biting Exclusive: deciding what criteria the study will not include; who does not fit into a specific population Inclusive: criteria that prospective participants must have to be considered eligible for a study

Experimental vs. Quasi experimental

Experimental: a method of research in the social sciences (such as sociology or psychology) in which a controlled experimental factor is subjected to special treatment for purposes of comparison with a factor kept constant. Control vs. experimental group Quasi-experimental: a type of design for testing an intervention in which participants are not randomly assigned to treatment conditions; also called a nonrandomized trial or a controlled trial without randomization

Rosenthal effect

If a researcher believes that their experiment is likely to results in a particular outcome, that bias will affect how the researcher conducts their work. The results will very likely sway towards the direction the researcher wanted, invalidating any study results "nonexperimental effect" interviewer's influence on respondents' answers

Temporal ambiguity

In a causal relationship, the cause precedes the effect. In RCTs, researchers create the independent variable and then observe the outcome, so establishing a temporal sequence is never a problem. In correlational studies, it may be unclear whether the independent variable preceded the dependent variable or vice versa.

Ways to increase internal & external validity of a study

Increase internal validity: single or double-blind study Single blind study: a type of clinical trial in which only the investigators know which treatment (or other intervention) the participants are receiving Double blind study: a situation (usually in a clinical trial) in which two study groups are blinded with respect to the groups that a study participant is in; often a situation in which neither the subjects not those who administer the treatment now who is in the experimental or control group Increase external validity: repetition and use of a diverse sample

Physiological measures

Involve the collection of physical data from subjects ex: wt, A1C, BMI Measures are objective and accurate Advantages: precision and accuracy Disadvantages: Expertise for using devices

Content validity

Items to represent the content Ways to measure → comparison w/ literature, panel of experts in subject area, test blueprint designed for content/level Degree to which an instrument has an adequate sample of items for the construct being measured Evaluated by expert evaluation, a quantitative measure - content validity index (CVI) - will also determine if number and type of questions are adequate

Sampling bias

Over-representing or under-representing population segment in terms of key characteristics Distortions that arise when a sample is not representative of the population from which it was drawn

selection bias

Reflects biases stemming from preexisting differences between groups. When people are not assigned randomly to groups, the groups being compared may not be equivalent

Validity of instrument

Quantity of variable of interest can be calculated w/ use of instrument Greater validity, more confidence in instrument Ways to measure validity → panel of experts, examination of literature Measurement may be w/ correlations - MUST HAVE reliability to have validity - face, content, construct, criterion

Characteristics of probability sampling

Random selection of elements from a population Each element in the population has an equal, independent chance of being selected (*not the same as random assignment like in RCTs) Only viable method of obtaining representative samples Allows researchers to estimate the magnitude of sampling error, which is the diff. b/w population values (e.g. the average of the population) and sample values (e.g. the average of the sample) Can be: Simple Random: Researchers establish a sampling frame― a list of population elements Ex: Out of 8709 Brazilian T2DM patients, randomly sampled 368 of those patients; all in a hat and draw, random number generator Stratified Random: The population is first divided into two or more strata, from which elements are randomly selected Ex: Split the population of Harding students into classification, sample random number of student each class - enhances representativeness Systematic: Involves the selection of every kth (ex. 10th, 5th etc.) case from a list, such as every 10th person on a list starting at a randomly selected number. Ex: Out of 1858 patients in a list, start at 294 and select every sixth patient Cluster Random: Used for large scale studies with far or spread out geographic sites/clusters. ***Associated with larger sampling error, but is considered more efficient*** - clusters become sampling units Ex. all nursing schools in southern region, randomize from each state

Criterion validity

Scores are correlated w/ external criterion Degree to which the instrument is r/t an external criterion Validity coefficient → calculated by analyzing the relationship between 2 scores on the instrument and the criterion 1) Concurrent validity: instrument's ability to distinguish individuals who differ on a present criterion 2) Predictive validity: instrument's ability to distinguish people whose performance differs on a future criterion Prediction of future behaviors Correlation coefficient closer to 1.00

Review types of instruments, scales, etc.

Self-report: a data collection method that involves a direct verbal report by participants (e.g. questionnaires, face-to-face interview, telephone interview) Evaluation: strong on directness, allows access to info otherwise not available to researchers; but can but can we be sure participants actually feel or act the way they say they do? - elicits socially acceptable answers, elicits answers desired by researcher, may not reflect true feelings or attitudes of subjects Interviews: a data collection method in which an interviewer ask questions of a respondent, either face-to-face, by telephone, or over the internet (e.g. Skype or FaceTime) Scale: a device that assigns a numeric score to people along a continuum; used to make fine quantitative discriminations among people with different attitudes, perceptions, & traits - Summated rating scales (composite scales): 5-7 choices and rate agreement/disagreement with subscales within that scale never/always - Attitude scales (Self report data collection instruments, respondents report attitudes or feelings, score given for the item responses, total may be obtained for individual subject or a group) - Likert Scale and Semantic Differential Scale Observational methods *excellent for capturing many clinical phenomena and behaviors* Time Sampling: sampling of time intervals for observation (random sample/systematic sample) Event Sampling: observation of integral events; requires researchers to either know when events will occur or wait for their occurrence to observe the response - Problem = Reactivity: behavioral distortions due to the known presence of an observer - Problem = observational bias: factors that can interfere with objective observation (cannot be eliminated, but can be minimized through careful observer training and assessment) The Flesch reading ease The Fog scale level The Flesch- Kincaid Grade Level Sources for research instruments: Mental measurement yearbooks Instruments for clinical health - care research The measurement of nursing outcomes health and psychosocial instruments

Semantic differential scale (Attitude Scale)

Subjects give their position or attitude about concept Continuum exists between 2 adjectives or phrases presented that are always opposites (e.g. happy to sad) Positions on continuum vary from 5-9 Scores then calculated for all the subjects - easier for subjects to use, adjectives used may be a problem

Correlation coefficient (pg. 235)

Summarizes the degree of relationship between variables ranging from -1 to +1 Higher the absolute value of the coefficient value of the coefficient (no matter that the sign), the stronger the relationship +1.00 means perfect positive relationship; both values go in the same direction Ex: as height increases, weight increases 0 means no relationships -1.00 means perfect negative relationship; variables inversely related Ex: higher levels of self-esteem have lower levels of depression

Hawthorne effect

The effect on the dependent variable resulting from subjects' awareness that they are participants under study. - respond in certain ways because know being watched

Internal validity

The extent to which it can be inferred that the independent variable is causing the outcome.

Validity

The measurement context is the degree to which an instrument is measuring the construct it purports to measure.

History threat

The occurrence of events concurrent with independent variable that can affect the outcome. Ex. studying the effectiveness of a senior center program to encourage flu shots among the elderly. Suppose a flu epidemic was aired in the national media at about the same time. The outcome variable, is now influenced by at least two forces and would be hard to disentangle the two effects.

Longitudinal design

a study design in which data are collected at more than one point in time, in contrast to a cross-sectional design

Cross-sectional design

a study design in which data are collected at one point in time; sometimes used to infer change over time when data are collected from different age or developmental groups

Prospective design

a study design that begins with an examination of a presumed cause (e.g. cigaretter smoking) and then goes forward in time to observe presumed effects (e.g. lung cancer); also called a cohort design

Retrospective design

a study design that begins with the manifestation of the outcome variable in the present (e.g. lung cancer), and a search for a presumed cause occuring in the past (e.g. cigarette smoking)

Response set bias

biases reflecting the tendency of some people to respond to items in characteristic ways, independently of item content - Social desirability: a bias in self-report instruments created when participants have a (tendency to misrepresent attitudes/traits by the participant giving answers that are socially acceptable) their opinions in the direction of answers consistent with prevailing social norms - Extreme: a bias in psychosocial scales created when participants select extreme response option (e.g. "strongly agree"), independent of the item's content (tendency to express extreme attitudes or extreme answers even if they don't have the extreme view) - Acquiescence: a bias in self-report instruments, especially in psychosocial scales, that occurs when participants characteristically agree with statements ("yea-sayers"), independent of content; (answer with ALL strongly agrees) Prevent: keep short and clear questions, avoid leading questions, avoid difficult or break up difficult concepts, good size sample

purpose of descriptive statistics

condense objective observations into usable data

Preexisting data

data is used from previous research - existing info is re-analyzed for new research - pt charts, records from agencies and organizations, personal docs, almanacs, professional journals

Direct observation

observing someone's hair color; a researcher can look at a person/subject and see the trait they are measuring. Physical characteristics, like hair color, eye color, and body type, are all able to be directly observed.

nonprobability sampling

does not involve selection of elements at random; is rarely representative of the population, will have sampling bias (most nrsg research is this type) - can be a limitation to a study

probability sampling

involves random selection of elements (each element has an equal, independent chance of being selected) - allows researchers to estimate the magnitude of sampling error (difference between population values & sample values), works to decrease sampling bias - stronger evidence, more for quantitative QUASI experimental studies

Bias

is any influence that distorts the results of a study and undermines validity Ex: age, gender, race, income

Multi-causality in nursing

is the portrayal of causality wherein several individual community and environmental factors may interact to cause a particular disease or condition or outcome.

Representatives of sample

one whose key characteristics closely approximate those of the population

Data analysis & Normal distributions

provides concrete view of more abstract concepts/propositions of the conceptual model or theoretical framework - inappropriate analysis -> inappropriate conclusion - level of measurement: great precision interval or ratio, rank things = ordinal scale, contrast things choose nominal scale - higher level of measurement/greater flexibility Normal distributions: - theoretical concept-data from repeated measures (interval/ratio) will group around a midpoint in a manner close to a normal curve - normal curve, mean, median, mode are equal = Bell curve - Skewness - samples not symmetrical; peak is off center - Positive skew: tail points to right (ex. personal income) - Negative skew: tail points to left (age of people with chronic illness)

Inclusion criteria

specifies characteristics of a population; characteristics that a prospective participant must have to be considered eligible for a study

Exclusion criteria

specifies characteristics that a population does not have

Strata

subpopulations of a population (ex: male & female)

Likert Scale (Attitude Scale)

summated rating scales - consist of several declarative statements (items) expressing viewpoints - responses are on an agree/disagree continuum (5-7 response options) - responses to items are summed to compute a total scale score Usually contains 5-7 responses for each item Responses range from strongly agree to strongly disagree Equal # of + or - worded items important Scores assigned to each response: Positive items scored in positive direction Negative items scored in negative/reverse direction Total scores are obtained

Threats to internal validity

temporal ambiguity, selection bias threat, history threat, maturation threat, mortality/attrition threat - instrumentation, setting, testing, lab setting, representativeness of sample, tight controls - hawthorne effect, rosenthal effect <- threats to external/internal validity

Measurement

the assignment of numbers to represent the amount of an attribute present in an object or person, using specific rules Advantages: removes guesswork, provides precise info, less vague than words Direct: BP, concrete Indirect: coping, self-care (use scale) - process of assigning numbers to variables - ways to assign numbers: count, rank, compare objects/events - quantification is the GOAL - numbers are assigned to data, concept applied with quantitative research designs - FOUR LEVELS (*NOIR*): nominal, ordinal, interval, ratio - a variable's level of measurement determines what mathematic operations can be performed in the statistical analysis (inferential is more precise) - errors of measurement: obtained score = true score +/- error - obtained score: an actual data values for a participant (anxiety scale score) - true score: the score that would be obtained with an infallible measure - error: the error of measurement caused by factors that distort measurement Two types of error: 1. systematic: variation/diff of score is consistently in the same direction; Ex. measuring wt, scale always weighs 2lb less than normal 2. random: difference in the pattern, not consistent; Ex. some it wt 2lb less or 3lb more Factors contribute to errors: situational contaminants (environment, hot/cold/loud), transitory personal factors, response set biases, item sampling

Reliability

the extent to which scores are free from measurement error. Test-retest reliability: replication takes the form of administering a measure to the same people on two occasions. The assumption is that for traits that have not changed, any differences in people's scores on the two testings are the result of measurement error. Internal consistency: In responding to a self-report item, people are influenced not only by the underlying construct but also by idiosyncratic reactions to words. By combining multiple items with various wordings, item irrelevancies to the extent that its items measure the same trait. For internal consistency, replication involves people's responses to multiple items during a single administrations. This captures the consistency across items.

Sample size

the number of study participants in the final sample - sample size adequacy is a key determinant of sample quality in quantitative research - sample size needs can and should be estimated through power analysis - the risk of "getting it wrong" (Statistical conclusion validity) increases when samples are too small

Sampling

the process of selecting a portion of the population to represent the entire population - *sample*

Probability sampling

the selection of sampling units (e.g. participants) from a population using random procedures (e.g. simple random sampling) Means of predicting Ex: 50% chance of rain Likelihood an event will occur Probability of getting heads with single flip of coin is 1 out of 2, or ½ or 0.5 Expressed as 50% or p=0.5

Quota sampling

type of nonprobability; a nonrandom sampling method in which "quotas" for certain subgroups, based on sample characteristics, are established to increase the representativeness of the sample - ID population strata and figuring out how many people are needed from each stratum

Purposive sampling

type of nonprobability; handpicking sample members - used in qualitative research mostly - other terms used: snowball, networking, "judgemental sampling"

Consecutive sampling

type of nonprobability; recruiting all people from an accessible population over a specific time interval

Convenience sampling

type of nonprobability; selecting the most conveniently available people as participants - MOST WIDELY USED APPROACH FOR QUANTITATIVE RESEARCHERS, IT IS THE MOST VULNERABLE TO SAMPLING BIAS - people who are readily available might be atypical of the population

Snowball sampling

type of nonprobability; the selection of participants through referrals from earlier participants (also called network sampling)

Random sampling

type of probability; selection of a sample such that each member of a population has an equal probability of being included